1. Field of the Invention
The present invention relates to an apparatus and method for cooling electronic components, and more particularly, an integrated cooling device for use with electronic components.
2. Description of Related Art
Machines and devices that change energy from one form (chemical, mechanical, thermal, or electrical) to another are rarely 100% efficient. Likewise, devices that change or modulate electrical energy in one form to electrical energy in another form are rarely 100% efficient. For example, a transistor, integrated circuit or a microprocessor may change direct current into alternating current or into pulses on many outputs that signify numerical data. Where direct current is changed into alternating current, only part of the direct current input power is changed into alternating current. The remainder of the input power appears as heat. Where direct current is changed into pulses signifying numerical data, all of the direct current input power becomes heat, and the output is information, not energy.
In many cases, the waste heat generated in a small machine or device is quite large, and the machine can neither radiate nor dissipate the heat to its surroundings, at a reasonable temperature, without providing an additional cooling mechanism. Semiconductors used in computers or as radio-frequency amplifiers are usually not very powerful, nor large, and do not generate a large amount of heat, but they may operate at a very high power density. The trend toward miniaturization and higher frequency operation of electronic devices, achieved through increased packing densities of gates in semiconductor devices, has resulted in ever increasing power densities. Predictions of powers of as much as 200 Watts over an area of two square centimeters (100 Watts per square centimeter) are being made. Such power densities are much too high for direct cooling with air.
There have been various attempts to address cooling issues associated with the increasing power densities of modern electronics. However, each of these attempted solutions has undesirable characteristics. A fairly basic approach to cooling electronic components is by free or forced convection of air over a “heat sink.” Typically a heat sink will include an array of cooling fins that collectively have a larger surface area than the electronic component to be cooled. These fins are attached to a thermally conductive base (or heat spreader). Heat from the electronic component is conducted through the conductive base to the roots of the cooling fins and along the fins toward their surfaces. The heat is then transferred from the cooling fins to the surrounding atmosphere via either free or forced convection.
The rate of heat transfer that can be accomplished by a heat sink is quite limited, however. The amount of power (the rate of heat flow) that can be transferred from the fins to the ambient air is a function of the average temperature difference between the fins and the air, the air velocity, and is proportional to the total fin area. For heat to spread radially outward to a large region, to which fins are attached, there must be a long heat path and a large temperature difference between the heated surface and the cool surface. (That is, in the case of a cooled electrical component, between the component itself at the contact point with the heat sink and the surfaces of the cooling fins). The total amount of heat transfer is proportional to the temperature difference between these two surfaces. At some power level, the device will become too hot to operate properly or perhaps even survive. This difficulty in effectively cooling electronic components is aggravated by the increased heat created by the increasing power densities encountered in modern electronic components.
In prior art, a solution to the problem of the large temperature difference in the “spreader” portion of a heat sink and the “fin” portion when used to cool high power-density components has been the replacement of the “heat spreader” with a “heat pipe.” In a “heat pipe,” the heat generated by a device evaporates a liquid. The vapor rises (perhaps through the central one of two coaxial pipes) and condenses on a cool area (perhaps the outer one of the coaxial pipes which passes through attached air-cooled fins). The liquid then returns to the heat-generating device so the process can repeat. Here, the heat transferred depends upon the mass flow rate of the liquid multiplied by the heat of vaporization of the liquid. Heat pipes with large ratios of cooled fin area to heated area can be made without using a large difference in temperature between the evaporation temperature (the device temperature) and the condensation (fin) temperature. Limitations on heat pipes include the onset of film boiling on the heated surface (which insulates the surface, often followed by rapid temperature rise and failure), pressure build-up which inhibits boiling within the heat pipe, and the requirement that the heat pipe be properly oriented with respect to gravity.
In order to make still larger reductions in the temperature of a cooled device with respect to ambient temperature, it is necessary to use forced convection of heat in solid or liquid material between the device and its surroundings. Some approaches to cooling heat-generating devices make use of moving solid structures to transport heat from one location to another. Such an approach can be more easily visualized as the heat transfer provided by a reciprocating piston in an internal combustion engine. A piston in an internal combustion engine is heated by the combustion of fuel during the combustion stroke of an Otto cycle engine and the piston then distributes this heat along the stroke depth of the cylinder walls as it travels.
The concept of using a moving part to transfer heat has been applied to the cooling of electronic components. Some devices using this transfer mechanism have used thermally conductive reciprocating sheets or rotating disks in thermal contact with a heat-generating component of a relatively smaller surface area. In these devices, heat is transferred from the heat-generating component to the moving part. Typically the movement of the part distributes the heat relatively evenly over the larger surface area of the moving part. Heat is then dissipated from the moving part through convective cooling (typically, with cooling fins).
Nevertheless, the use of a moving part to remove heat from an electronic component, as provided by the prior art, has several shortcomings. First, in some of the prior-art moving-part cooling devices, the moving part is in direct contact with the heat-generating device. This direct contact leads to friction heating and wear of both the moving part and the heat-generating component. If, as in other prior-art devices, a fluid is used to provide a lubricating interface between the heat-generating component and the moving part, heat must be transferred between the heat-generating component, the relatively poorly conductive lubricating fluid, and into the moving part. In some prior art moving-part-based cooling devices, the heat would then be transferred through another poorly-conductive-fluid interface to cooling fins. Therefore, the heat transfer path provided by a moving-part-based cooling device is impeded as it passes through several interfaces with low thermal conductivity.
There is another reason that the use of a moving solid part to transfer heat is inferior as compared to simple liquid cooling. The solid moving part used to transfer heat is often a thermally conductive material such as a metal. The thermal capacity of a metal (i.e., the specific heat multiplied by density and temperature rise, or the amount of energy that may be carried by a given volume of the material) is often less than that of a liquid such as water or many other liquids with similar properties. Therefore, the use of a liquid as a heat transfer medium is preferable to a solid moving part because additional heat may be carried by an equal volume of liquid. While still liquids and liquids in laminar flow tend to have lower thermal conductivities (which limit the liquid's ability to rapidly distribute heat energy) than thermally conductive materials such as metals, a liquid in the turbulent flow regime will have similar heat distributing abilities.
An additional advantage of liquid cooling over solid moving-part-based cooling is that the cooling effect provided by a liquid cooling medium can be significantly enhanced through the use of nucleate boiling. Nucleate boiling occurs in liquid cooling devices with high-velocity cooling streams when the temperature of the heat-generating component is slightly higher than the boiling temperature of the cooling liquid. In nucleate boiling, very small bubbles of cooling liquid vapor are swept off of the interface between the heat generating component and the cooling liquid by the flowing liquid. These bubbles then condense within the cooling liquid stream. Through the nucleate boiling mechanism, the heat transfer coefficient (a measure of the cooling ability of the system) may be increased by a factor of as much as ten over normal liquid cooling. The enhanced cooling ability provided by nucleate boiling could not occur in a cooling system relying on a solid moving part for heat transfer. High-velocity fluid flow inhibits the onset of film boiling and consequent “burn out” that can occur in low-velocity boiling-fluid systems (as mentioned above in connection with heat pipes). High-velocity water-cooling systems utilizing nucleate boiling have reliably transferred as much as several kilowatts of power per square centimeter of cooled area.
There have been many prior art approaches to cooling heat-generating components using liquid cooling. In many common applications of liquid cooling, the heat generated by a device is transferred to a high-velocity liquid (typically flowing through the tiny channels of a heat sink in contact with the device). The liquid is conveyed (typically in another hose, pipe, or other conduit) to a small auto-radiator-like heat exchanger, and finally, the liquid is returned (typically via hose, pipe, or other conduit) to the pump and the heat-generating device so the process can repeat. In some cases, by allowing direct contact between the cooling fluid and the semiconductor package, the transfer of heat is enhanced. This results from the elimination of temperature differences between the heat sink and fluid and in the heat-sink compound used in the joint between the component and the heat sink. Here, the heat transferred again depends upon the mass flow rate of the liquid multiplied, in this case, by the specific heat of the liquid and the temperature change. For maximum heat transfer, the recirculating cooling liquid must be pumped, at high velocity, through small channels in the heat exchanger, and connecting pipes. Especially when operating at the high mass-flow rates required for maximum heat transfer, such systems may suffer from large pressure drops, or “head losses.”
Another approach to using liquid cooling to cool heat generating components attempts to address the shortcomings of cooling systems using recirculating liquid cooling. In this alternate approach, the cooling components are more integrated. A cooling device housing is thermally connected to a heat generating component, and a cooling liquid is contained in the housing and is circulated inside the housing by an impeller. Completely integrated cooling devices are more easily usable in high power density electronic devices such as computers. The application is simpler because the cooling device is essentially a single component that thermally contacts the heat generating electronic components (rather than using separate liquid conduits, pumps and heat exchangers that would be required for a recirculating liquid cooling system). Additionally, by integrating the cooling components, the pump energy required can be used directly to transport the cooling liquid at high velocity across the heated surface rather than to overcome the head losses that are encountered in a less integrated liquid cooling system.
But, prior art attempts to integrate cooling device components have been problematic. Since prior-art integrated cooling devices have thermally or electrically conductive impellers driven by externally-created electromagnetic fields, heat transfer must occur between the impeller and the cooling fluid with its relatively poor thermal conductivity. A more direct heat transfer mechanism could provide enhanced cooling. Further, the thermally and electrically conductive impeller of the prior art requires compromises in the housing of the cooling device. The prior art integrated cooling devices have a composite housing, including layers of metal and plastic, to allow external electromagnetic fields to pass through the surface of the housing and motivate the impeller. The composite structure may not be able to withstand the amounts of heat generated by ever increasing power densities of modern electronic components. This indirectly motivated impeller of the prior art may also revolve at a lower speed than the speed of the motivating electromagnetic field. Therefore, another drawback of the prior art is poor liquid circulation and possible difficulties in achieving a turbulent flow in the cooling liquid. Alternatively, the impeller of prior-art integrated cooling devices is motivated by a shaft that, in turn, is driven by an external motor. Such an external motor shaft drive arrangement requires a seal between the motor and the housing and impeller. As the seal begins to wear, leaking cooling liquid can lead to problems such as leakage, reduced cooling effectiveness and ultimately damage the heat-generating components or adjacent electronic components Additionally, the prior-art attempts to provide an integrated liquid cooling mechanism have failed to take advantage of the additional cooling ability provided by nucleate boiling.
Therefore, in light of the prior art, there is a need for an integrated cooling device that has a direct, high thermal capacity, low-temperature-difference heat-transfer path, that circulates the cooling liquid in turbulent flow without requiring sealing between rotating components and the housing, and that can advantageously use nucleate boiling to cool the heat-generating components.